Classification of Metacognitive into Two Catagories to Support the Learning Process

(1)

Article received 11/9/2016; Approved 10/1/2017

Classification of Metacognitive into Two Catagories to Support the Learning Process

Husnul Rahmawati Sakinnah, Adhistya Erna Permanasari, Indah Soesanti Department of Electrical and Information Technology-Universitas Gadjah Mada

E-mail: husnul.ti14@mail.ugm.ac.id

Abstract: Learning outcomes are the patterns of actions, values, understanding, attitudes, appreciation and skills. Learning outcomes are related to the metacognitive of student where the elements contained in metacognitive is cognitive. The relationship between cognitive and metacognitive which is the foundation of cognitive is metacognitive. There are two components such as knowledge of metacognitive and regulation of metacognitive. In the learning process, cognitive factors are not the only one that can support, but also a metacognitive factor is a very influential factor for the success of the learning process. Thus, it is very important to do with a deeper analysis about metacognitive by identifying metacognitive level to support the learning process. Identification of metacognitive is performed by using Naïve Bayes Classifier algorithm (NBC) which NBC is one of an algorithm that is used for classification algorithm for data mining. In these studies, it is obtained that the accuracy scored 88,0597% when tested using NBC.

Key Words: metacognitive, knowledge of metacognitive, regulation of metacognitive, cognitive, Naïve Bayes Classifier (NBC)

Abstrak: Hasil pembelajaran merupakan pola tindakan, nilai-nilai, pemahaman, sikap, apresiasi dan keterampilan. hasil belajar terkait dengan metakognitif siswa di mana unsur-unsur yang terkandung dalam metakognitif adalah kognitif. Hubungan antara kognitif dan metakognitif merupakan dasar dari kognitif adalah metakognitif. Terdapat dua komponen dalam pengetahuan metakognitif dan regulasi metakognitif. Dalam proses pembelajaran, faktor kognitif bukan satu-satunya yang dapat mendukung, tetapi juga faktor metakognitif adalah faktor yang sangat berpengaruh bagi keberhasilan proses pembelajaran. Jadi, sangat penting untuk melakukan analisis yang lebih mendalam tentang metakognitif dengan mengidentifikasi tingkat metakognitif untuk mendukung proses pembelajaran. Identifikasi metakognitif dilakukan dengan menggunakan algoritma Naïve Bayes Classifier (NBC) dimana NBC merupakan salah satu algoritma yang digunakan untuk algoritma klasifikasi untuk data mining. Dalam penelitian tersebut diperoleh bahwa nilai akurasi adalah 88,0597% saat diuji menggunakan NBC.

Kata kunci: Metakognitif, Pengetahuan Metakognitif, Peraturan Metakognitif, Kognitif, Naïve Bayes Classifier (NBC)

Volume 5, Number 1, March 2017, pp. 11–16

C

ognitive development is a determinant of learners intellectual intelligence which is the cognitive ability will continue to evolve with physical factors and brain as the main of it. In addition, the development of cognitive followed with how to organize or manage the ability to solve a problem. In this case, cognitive of someone is unable to walk solitary but they need to be controlled and set up so if someone is going to use their own cognitive, they need the ability to organize the cognitive. Learning outcomes

are related with metacognitive of student where the elements contained in metacognitive is cognitive.

Relation between metacognitive and cognitive is metacognitive means a foundation of cognitive. The successive of learning outcomes of students is determined by cognitive and metacognitive factors as a support and foundation for cognitive learners.

Someone must have awareness about their mind and be able to manage and controlled it. In this case the meaning of ability mentioned before is called as

11

(2)

metacognitive. Metacognitive ability sometimes represented by aiming their own selves. As a simply, metacognitive defined as “thinking about thinking”

(Elawar, 1995). Basically, metacognitive is an awareness of thinking about what is known and what is unknown. In the learning context, learner know how to learn, knowing their capabilities and learning modality and determine the best learning strategies for effective learning.

Metacognition is an object of cognition which is classified from learner experience and information transform to perform monitoring and metacognition control, learners are actively involved in how to learn and factors that guarantee their learning process (Winne, . In the learning process, there is a learning evaluation, in this case, role of metacognitive is very influential for them so learners can cultivate the ability of self-learning evaluation learning that has been passed which in principle is considered as an important component of self-learning (McCabe, 2011).

This study focuses on extracting the aspects that exist in metacognitive and gain the learners’ knowledge that will be used in identifying metacognitive level.

There are two main components in metacognitive that is knowledge of metacognitive and regulation of metacognitive (Sware, 1994). Knowledge of metacognitive is an indicator of how well for someone to use the method and strategies to control and enhance learning and knowledge (McCabe, 2011;

Nett, et al, 2012; Kelemen, et al, 2000). Here is a map of research shown in Figure 1.

Figure 1. Research Scheme

Previous research about image classifier to classify the mammogram images by classify into malignant, benign and normal cases. Correlation based feature selection (CFS) to reduce 50% of features.

Correlation based feature selection (CFS) has a function to remove the uncorrelated features and select the significant feature from the original features (Choridah, 2014).

As a simply, metacognitive defined as “thinking about thinking” (Coskun, 2010). Metacognition is a reflection of the mind, thought to their own mind. Nett, et al stated that metacognitive strategy is a strategy that involves cognitive learning and motivation, metacognitive can be interpreted as a theoretical model which has an important role in the success of independent learning (Sharabiani, 2014; Kelemen, 2000;

Durall, 2012; Biswas, 2014). Basically, Metacogntive is the awareness of what is known and what is unknown. In the context of learning, students know how to learn, determine the ability of learning, learning of the owned modality and determine the best learning strategies for effective learning. Flavel defined metacognitive as one’s knowledge with respect to the cognitive process and product by itself or everything which is related to the process and the product (John, 2006).

Metacognitive is a knowledge about the regulation of cognitive activity of individuals in the learning process (Veenmann, 2006). Corebima states that metacognition is related to think how to think, knowing what we know, and we do not know, learn how to learn, develop thought process on an ongoing basis so it can be used to solving the problem (Corebima, 1994). Schraw devide metacognitive into two main components i.e metacognitive knowledge and metacognitive control. All components can be seen in Table 1.

As stated in the National Education System Re- gulation. Education is a conscious and deliberate effort to create an atmosphere of learning and the learning process so learners are actively developing their poten- tial to have spiritual power of religion, self-control, personality, intelligence, noble character and skills which are needed for them, society for the nation and state. In a real practice, there is a gap between reality and the aim of education in which the results of empiri- cal observations indicate that most learner’s failures in learning process not due to cognitive factors but metacognitive factors, because basically students who entered on a favorite school have a background in an excellent cognitive, however there are some failures in learning, so the evaluation to identify the metacognitive level is one of solution to improving the quality of education.

METHOD

In recent years, an analysis of student’s metacognition is very important and a lot of research topics being work on by some researches (Kelemen,

(3)

Table 1. Componets of Metacognitive

2000; Xu, 2009; McCabe, 2011. Previous research using cognitive and metacognitive to analyze student learning behaviors (Biswas, 2011), metacognitive interaction is monitored according to learning process (Schwartz and Chase, 2009) strategy is used to analyze an Open Ended-Learning Environments (Biswas, 2000).

This research was done in 8 stages, shown in figure 2. The first stage is the study of the theory which is used to compose a research instrument that will be used for data retrieval. Schraw and Dennison makes absolutely an instrument to measure the components of metacognition called Metacognitive Awareness Inventory (Inventory of Metacognition) (Sware, 1994). That inventory contains knowledge of metacognitive and control metacognitive factors.

There are 17 questions related to knowledge factor metacognition and there are 35 questions related to factor settings or metacognition control.

The research began with collecting the data using MAI. Then 52 items of MAI selected using feature selection. Feature selection is a process of selecting appropriate subset of original featured. Optimization of feature selection from the subset is a measure of evaluation criteria (Liu, 2012). On this research of 52 items which is divides into 8 indicators on each indicator will be grouped into two main components.

Sub attribute of 8 indicators are shown in Table 1, included Declarative Knowledge (ZDK), Procedural Knowledge (ZPK), Conditional

Figure 2. Research Methodology

Knowledge (ZCK), Planning (ZPN), Information Management Strategies (ZIMS), Comprehension Monitoring (ZCM), Debugging Strategies (ZDS) and Evaluation (ZEV).

This research uses a z-score analysis to compare one attribute with another attribute and will be classified into two components. Z-score also known as zero-mean, where the value of attribute of A normalized based on average value and standard deviation of attribute of A. A value v of A has normalized

v^′ = v − ̅

(1) Component Indicator Description

Knowledge of metacognitive

Declarative knowledge

Knowledge of the skills,

resources and

capabilities of a person as a learner

Procedural knowledge

Knowledge of how to implement learning steps

Conditional knowledge

Knowledge of why and when learning steps used

Regulation of metacognitive

Planning Plan learning process, setting the goals, allocating the resources for learning priorities.

Information management

Skills and strategy to efficiently process information (such as organizing, elaboration, summarizing, focusing election)

Monitoring Assessment of learning or use of strategies Debugging The use of strategies to

improve/check for errors in the understanding and implementation/achieve ment

Evaluation Analyzing the achievements and effectiveness of the strategy after learning process

(4)

attribute become v’ where _A and _A are mean and standard deviation (Junaedi, et al, 20111). The equation z-score can be seen in equation 1.

A NBC (Naïve Bayes Classifier) algorithm used to classify the scored results from z-score process.

Naïve bayes classifier is classification model which has several advantages, including: 1) easy to learn and understand, 2) more efficient, 3) have high accuracy (Mark & Sucar, 2006). NBC is a method of probability and statistics stated by the UK scientist Thomas Bayes. Bayes theorem can predict future probability based on previous experience. This theorem compared with the Naïve which is assumed that every attribute is independent (Mark & Sucar, 2006). Here are the equation from theorem naïve bayes as shown by equation (2) and (3).

posterior = likelihood x prior evidence p(y|x) = p(x, y)

p(x) = p(x|y) ( )

∑ ₌₁ ( | ^′)p(y′)

NBC algorithm is commonly used for classification data set by using a generative function. The classification process requires a number of dues to determine best suited class for analyzed samples. In addition, NBC concept can be used to correlate the hypothesis too. Correlation of hypothesis is shown P(Y|X) = ^{P(Y) ∏} ^P(Xⁱ^|Y)

q i=1

P(X)

on the label of a mapping class that will be used in the classification. The input vector that contain the features in the concept of NBC is X and class label is Y, so Naïve Bayes can be written by . The notation on Naïve Bayes showed by posterior prob- ability which represented by Y and prior probability which represented by X. Therefore, Bayes theorem can be seen in equation (4) (Bustami, 2014).

Learning outcomes is not only affected by cognitive factor but there is metacognition factors that supporting the learning process. Analyzing this research done by classify knowledge of metacognitive and regulation of metacognitive (Sware, 1994).

This research observed first grade of computer network engineering in SMKN 2 Singosari, Malang as a population and use two classes of it as a sample.

The metacognitive questionnaires were distributed to 67 students. The goal of this research is to shown the dominant component between knowledge of metacognitive or regulation of metacognitive by using combination of NBC and z-score analysis.

RESULTS AND DISCUSSION

The result of feature selection will be scoring values for each attribute in each z-score. Results of feature selection using z-score are shown in table 2.

Scoring is done by sorting based on highest value at each scoring of respondents where the sample is shown in table 3. In table 3 are shown 10 sample of data from 67 data. Z-score has a function to balance between two main components, because in one

Table 2. Feature Selection Using ZScore

ZDK ZPK ZCK ZPN ZIMS ZCM ZDS ZEV

0.304 -0.363 -0.041 0.234 0.050 -0.398 15.152 10.718 -0.787 -0.363 -0.846

-

10.176 -0.278 -

11.428 -0.384 -0.278 -0.132 -1.046 -0.846 0.025 -0.442 -0.398 0.566 -0.548 -0.569 -0.704 -0.577

-

10.176 -0.771 -0.398 -0.700 -0.548 -0.569 -0.363 -0.041 -0.600 0.706 -0.647 -0.067 -0.278 -0.351

-

13.880 -0.309 -0.392 -0.114 -0.895 -0.700 -0.278 16.136 0.321 0.228 21.111 20.184 18.344 0.566 2.152

-

18.782 -0.704 -

21.878 -0.600 -0.771 - 11.428

- 16.495

- 21.682 0.304 0.663 -0.309 0.442 0.706 -0.647

-

13.330 -0.008 -0.787 -0.021

-

16.510 -1.435 -1.591 -0.398 -

16.495 -1.088 (2)

(3)

(4)

(5)

Table 3. Sample of Scoring Attribute

Table 4. Classification By Using Naïve Bayes Classifier

Correctly Classified Instances

58 88.0597%

Incorrectly Classified Instances

8 11.9403%

Kappa statistic 0.7599 Mean absolute error 0.146 Root mean squared error 0.2688 Relative absolute error 29.3327%

Root relative squared error 53.861%

Total Number of Instances 67%

Attribute Class

RK KC

(0,54) (0,46) ZDK

Mean 4.3333 5.0645

Std.dev 2.3214 2.0468

Weight sum 36 31

Precision 1 1

ZPK

Mean 4.7222 4.5484

Std.dev 2.1809 2.2409

Weight sum 36 31

Precision 1 1

ZCK

Mean 4.7778 4.1935

Std.dev 2.0831 2.0386

Weight sum 36 31

Precision 1 1

ZPN

Mean 4.4444 4.2581

Std.dev 2.1402 2.1845

Weight sum 36 31

Precision 1 1

ZIMS

Mean 3.7778 4.9355

Std.dev 2.225 2.5645

Weight sum 36 31

Precision 1 1

ZCM

Mean 3.9722 4.6129

Std.dev 2.2421 2.4053

Weight sum 36 31

Precision 1 1

ZDS

Mean 6.4722 2.6452

Std.dev 1.6242 1.8413

Weight sum 36 31

Precision 1 1

ZEV

Mean 3.3333 5.7419

Std.dev 1.9861 1.5651

Weight sum 36 31

Precision 1 1

ZDK ZPK ZCK ZPN ZIMS ZCM ZDS ZEV CLASS

6 2 1 5 4 3 8 7 RK

4 6 3 2 7 1 5 8 KC

6 1 2 7 4 5 8 3 RK

6 3 5 1 2 8 4 7 KC

3 4 7 2 2 1 6 5 RK

5 1 6 4 8 2 3 7 RK

5 2 1 8 7 6 3 4 KC

2 7 8 5 6 4 3 1 RK

5 7 3 6 8 2 1 4 RK

6 8 1 4 3 7 2 5 KC

component has 3 indicators such as in knowledge metacognitive there are declarative knowledge (DK), procedural knowledge (PK) and conditional knowledge (CK). And for regulation of metacognitive, there are 5 indicator such as planning (PN), information management (IMS), Monitoring (CM), Debugging (DS) and Evaluation (EV). Analyzing using zscore can make it easy to analyze the diffirent sum of each attribute.

Based on result from analysis using Naive Bayes Classifier, obtained an accuracy 88.0597% with correctly classified instances is 58. While incorrectly classified instance is 8 with 11.9403% errors.

CONCLUSIONS

Classification of metacognition is done by using feature selection and using z-score approach to give scoring in each attribute because scoring indicators on each component is different. Then, to balance the assessment with the sum of different indicators can prove the solution for data that has a number of different attribute. In these studies obtained accuracy 88,0597% when tested using NBC.

The results of this research is knowledge of cognition has percentage is 50% and the percentage

(6)

of knowledge of regulation is 48%. So in this case, most of the students have knowledge of cognition more dominant than regulation of knowledge.

REFERENCES

Biswas, G. (2000). Analyzing Students Metacognitive Strategies in Open-Ended Learning Environments, 209–214.

Biswas, G., Kinnebrew, J. S., & Segedy, J. R. (2014). Using a Cognitive/Metacognitive Task Model to Analyze Students Learning Behaviors, 190–201.

Cardelle-Elawar, M. (1995). Effects of metacognitive instruction on low achievers in mathematics prob- lems. Teaching and Teacher Education, 11(1), 81–

95. http://doi.org/10.1016/0742-051X(94)00019-3 Choridah, L. (2014). Based on Texture Feature and

Correlation based Feature Selection.

Corebima, A. D. (1997). Metacognitive Skill Measurement Integrated In Achievement Test The Metacognitive Awareness Inventory ( MAI ) was developed by Schraw and Dennison (1994) to assess metacognitive knowledge and metacognitive regulation which they referred to as the SM31050, (1994).

Durall, E., & Analysis, N. (2012). Learning Analytics as a Metacognitive Tool Fostering Self-directed, 380–

384.

John, F. (n.d.). John Flavell Metacognition Theory.

Junaedi, H., Budianto, H., & Maryati, I. (2011). Data transformation pada data mining, 7, 93–99.

Kelemen, W. L., Frost, P. J., & Weaver, C. a. (2000). Individ- ual differences in metacognition: evidence against a general metacognitive ability. Memory & Cog- nition, 28(1), 92–107. http://doi.org/10.3758/

BF03211579

Liu, H., Member, S., Yu, L., & Member, S. (2005). Algorithms for Classification and Clustering, 17(4), 491–502.

Mart, M., & Sucar, L. E. (2006). Learning an Optimal Naive Bayes Classifier, 18–21.

McCabe, J. (2011). Metacognitive awareness of learning strategies in undergraduates. Memory & Cogni- tion, 39, 462–476. http://doi.org/10.3758/s13421- 010-0035-2

Nett, U. E., Goetz, T., Hall, N. C., & Frenzel, A. C. (2012).

Metacognitive Strategies and Test Performance:

An Experience Sampling Analysis of Students’

Learning Behavior. Education Research Interna- tional, 2012, 1–16. http://doi.org/10.1155/2012/

958319

Schraw, G. and D. R. . (1994). Assessing metacognitive awareness.

Schwartz, D., & Chase, C. (2009). Interactive metacognition: Monitoring and regulating a teachable agent. Handbook of Metacognition in Education, (650), 340–358.

Sharabiani, A., Karim, F., Sharabiani, A., Atanasov, M., &

Darabi, H. (2014). An Enhanced Bayesian Network Model for Prediction of Students’ Academic Perfor- mance in Engineering Programs, (April), 832–837.

Teknik, D., Universitas, I., Bayes, N., Bayes, N., & Bayes, N. (n.d.). Penerapan Algoritma Naive Bayes Untuk Mengklasifikasi Data Nasabah.

The Effect Of Metacognitive Strategy Training On The.

(2010), 4(1), 35–50.

Veenman, M. V. J., Hout-Wolters, B. H. a. M., & Afflerbach, P. (2006). Metacognition and learning: conceptual and methodological considerations. Metacog- nition and Learning, 1(1), 3–14. http://doi.org/

10.1007/s11409-006-6893-0

Winne, P. H., & Baker, R. S. J. d. (2013). The Potentials of Educational Data Mining for Researching Meta- cognition, Motivation and Self-Regulated Learn- ing. JEDM - Journal of Educational Data Mining, 5(1), 1–8.

Xu, X. (2009). Evaluation of the relationship between one’s differentiation of cognitive processes and meta- cognitive self-ratings. Behavior Research Meth- ods, 41(1), 244–255. http://doi.org/10.3758/

BRM.41.1.244